1 // Copyright (c) 2011 The Chromium Authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file.
5 // This file contains the SdchManager class and two nested classes
6 // (Dictionary, DictionarySet). SdchManager::Dictionary contains all
7 // of the information about an SDCH dictionary. The manager is
8 // responsible for storing those dictionaries, and provides access to
9 // them through DictionarySet objects. A DictionarySet is an object
10 // whose lifetime is under the control of the consumer. It is a
11 // reference to a set of dictionaries, and guarantees that none of
12 // those dictionaries will be destroyed while the DictionarySet
13 // reference is alive.
15 #ifndef NET_BASE_SDCH_MANAGER_H_
16 #define NET_BASE_SDCH_MANAGER_H_
23 #include "base/gtest_prod_util.h"
24 #include "base/memory/ref_counted.h"
25 #include "base/memory/scoped_ptr.h"
26 #include "base/memory/weak_ptr.h"
27 #include "base/observer_list.h"
28 #include "base/threading/thread_checker.h"
29 #include "base/time/time.h"
30 #include "net/base/net_export.h"
31 #include "net/base/sdch_problem_codes.h"
43 // Provides global database of differential decompression dictionaries for the
44 // SDCH filter (processes sdch enconded content).
46 // The SdchManager maintains a collection of memory resident dictionaries. It
47 // can find a dictionary (based on a server specification of a hash), store a
48 // dictionary, and make judgements about what URLs can use, set, etc. a
51 // These dictionaries are acquired over the net, and include a header
52 // (containing metadata) as well as a VCDIFF dictionary (for use by a VCDIFF
53 // module) to decompress data.
55 // A dictionary held by the manager may nonetheless outlive the manager if
56 // a DictionarySet object refers to it; see below.
57 class NET_EXPORT SdchManager
{
60 typedef std::map
<std::string
, scoped_refptr
<base::RefCountedData
<Dictionary
>>>
63 class NET_EXPORT_PRIVATE Dictionary
{
65 // Construct a vc-diff usable dictionary from the dictionary_text starting
66 // at the given offset. The supplied client_hash should be used to
67 // advertise the dictionary's availability relative to the suppplied URL.
68 Dictionary(const std::string
& dictionary_text
,
70 const std::string
& client_hash
,
71 const std::string
& server_hash
,
73 const std::string
& domain
,
74 const std::string
& path
,
75 const base::Time
& expiration
,
76 const std::set
<int>& ports
);
80 // Sdch filters can get our text to use in decoding compressed data.
81 const std::string
& text() const { return text_
; }
83 const GURL
& url() const { return url_
; }
84 const std::string
& client_hash() const { return client_hash_
; }
85 const std::string
& server_hash() const { return server_hash_
; }
86 const std::string
& domain() const { return domain_
; }
87 const std::string
& path() const { return path_
; }
88 const base::Time
& expiration() const { return expiration_
; }
89 const std::set
<int>& ports() const { return ports_
; }
91 // Security methods to check if we can establish a new dictionary with the
92 // given data, that arrived in response to get of dictionary_url.
93 static SdchProblemCode
CanSet(const std::string
& domain
,
94 const std::string
& path
,
95 const std::set
<int>& ports
,
96 const GURL
& dictionary_url
);
98 // Security method to check if we can use a dictionary to decompress a
99 // target that arrived with a reference to this dictionary.
100 SdchProblemCode
CanUse(const GURL
& referring_url
) const;
102 // Compare paths to see if they "match" for dictionary use.
103 static bool PathMatch(const std::string
& path
,
104 const std::string
& restriction
);
106 // Compare domains to see if the "match" for dictionary use.
107 static bool DomainMatch(const GURL
& url
, const std::string
& restriction
);
109 // Is this dictionary expired?
110 bool Expired() const;
112 void SetClockForTesting(scoped_ptr
<base::Clock
> clock
);
115 friend class base::RefCountedData
<Dictionary
>;
117 // Private copy-constructor to support RefCountedData<>, which requires
118 // that an object stored in it be either DefaultConstructible or
120 Dictionary(const Dictionary
& rhs
);
122 // The actual text of the dictionary.
125 // Part of the hash of text_ that the client uses to advertise the fact that
126 // it has a specific dictionary pre-cached.
127 std::string client_hash_
;
129 // Part of the hash of text_ that the server uses to identify the
130 // dictionary it wants used for decoding.
131 std::string server_hash_
;
133 // The GURL that arrived with the text_ in a URL request to specify where
134 // this dictionary may be used.
137 // Metadate "headers" in before dictionary text contained the following:
138 // Each dictionary payload consists of several headers, followed by the text
139 // of the dictionary. The following are the known headers.
140 const std::string domain_
;
141 const std::string path_
;
142 const base::Time expiration_
; // Implied by max-age.
143 const std::set
<int> ports_
;
145 scoped_ptr
<base::Clock
> clock_
;
147 void operator=(const Dictionary
&) = delete;
150 // A handle for one or more dictionaries which will keep the dictionaries
151 // alive and accessible for the handle's lifetime.
152 class NET_EXPORT_PRIVATE DictionarySet
{
156 // Return a comma separated list of client hashes.
157 std::string
GetDictionaryClientHashList() const;
159 // Lookup a given dictionary based on server hash. Returned pointer
160 // is guaranteed to be valid for the lifetime of the DictionarySet.
161 // Returns NULL if hash is not a valid server hash for a dictionary
162 // named by DictionarySet.
163 const SdchManager::Dictionary
* GetDictionary(const std::string
& hash
) const;
168 // A DictionarySet may only be constructed by the SdchManager.
169 friend class SdchManager
;
172 void AddDictionary(const std::string
& server_hash
,
173 const scoped_refptr
<base::RefCountedData
<
174 SdchManager::Dictionary
>>& dictionary
);
176 DictionaryMap dictionaries_
;
178 DISALLOW_COPY_AND_ASSIGN(DictionarySet
);
184 // Clear data (for browser data removal).
187 // Record stats on various errors.
188 static void SdchErrorRecovery(SdchProblemCode problem
);
190 // Enables or disables SDCH compression.
191 static void EnableSdchSupport(bool enabled
);
193 static bool sdch_enabled() { return g_sdch_enabled_
; }
195 // Enables or disables SDCH compression over secure connection.
196 static void EnableSecureSchemeSupport(bool enabled
);
198 static bool secure_scheme_supported() { return g_secure_scheme_supported_
; }
200 // Briefly prevent further advertising of SDCH on this domain (if SDCH is
201 // enabled). After enough calls to IsInSupportedDomain() the blacklisting
202 // will be removed. Additional blacklists take exponentially more calls
203 // to IsInSupportedDomain() before the blacklisting is undone.
204 // Used when filter errors are found from a given domain, but it is plausible
205 // that the cause is temporary (such as application startup, where cached
206 // entries are used, but a dictionary is not yet loaded).
207 void BlacklistDomain(const GURL
& url
, SdchProblemCode blacklist_reason
);
209 // Used when SEVERE filter errors are found from a given domain, to prevent
210 // further use of SDCH on that domain.
211 void BlacklistDomainForever(const GURL
& url
,
212 SdchProblemCode blacklist_reason
);
214 // Unit test only, this function resets enabling of sdch, and clears the
216 void ClearBlacklistings();
218 // Unit test only, this function resets the blacklisting count for a domain.
219 void ClearDomainBlacklisting(const std::string
& domain
);
221 // Unit test only: indicate how many more times a domain will be blacklisted.
222 int BlackListDomainCount(const std::string
& domain
);
224 // Unit test only: Indicate what current blacklist increment is for a domain.
225 int BlacklistDomainExponential(const std::string
& domain
);
227 // Check to see if SDCH is enabled (globally), and the given URL is in a
228 // supported domain (i.e., not blacklisted, and either the specific supported
229 // domain, or all domains were assumed supported). If it is blacklist, reduce
230 // by 1 the number of times it will be reported as blacklisted.
231 SdchProblemCode
IsInSupportedDomain(const GURL
& url
);
233 // Send out appropriate events notifying observers that a Get-Dictionary
234 // header has been seen.
235 SdchProblemCode
OnGetDictionary(const GURL
& request_url
,
236 const GURL
& dictionary_url
);
238 // Send out appropriate events notifying observers that a dictionary
239 // was successfully used to decode a request. Note that this can happen
240 // after a dictionary has been deleted from the SdchManager (because
241 // DictionarySets retain references to deleted dictionaries).
242 void OnDictionaryUsed(const std::string
& server_hash
);
244 // Get a handle to the available dictionaries that might be used
245 // for encoding responses for the given URL. The return set will not
246 // include expired dictionaries. If no dictionaries
247 // are appropriate to use with the target_url, NULL is returned.
248 scoped_ptr
<DictionarySet
> GetDictionarySet(const GURL
& target_url
);
250 // Get a handle to a specific dictionary, by its server hash, confirming
251 // that that specific dictionary is appropriate to use with |target_url|.
252 // Expired dictionaries will be returned. If no dictionary with that
253 // hash exists that is usable with |target_url|, NULL is returned.
254 // If there is a usability problem, |*error_code| is set to the
255 // appropriate problem code.
256 scoped_ptr
<DictionarySet
> GetDictionarySetByHash(
257 const GURL
& target_url
,
258 const std::string
& server_hash
,
259 SdchProblemCode
* problem_code
);
261 // Construct the pair of hashes for client and server to identify an SDCH
262 // dictionary. This is only made public to facilitate unit testing, but is
264 static void GenerateHash(const std::string
& dictionary_text
,
265 std::string
* client_hash
, std::string
* server_hash
);
267 // For Latency testing only, we need to know if we've succeeded in doing a
268 // round trip before starting our comparative tests. If ever we encounter
269 // problems with SDCH, we opt-out of the test unless/until we perform a
270 // complete SDCH decoding.
271 bool AllowLatencyExperiment(const GURL
& url
) const;
273 void SetAllowLatencyExperiment(const GURL
& url
, bool enable
);
275 base::Value
* SdchInfoToValue() const;
277 // Add an SDCH dictionary to our list of availible
278 // dictionaries. This addition will fail if addition is illegal
279 // (data in the dictionary is not acceptable from the
280 // dictionary_url; dictionary already added, etc.).
281 // If |server_hash| is non-null, returns the server hash that may be
282 // used as an argument to GetDictionarySetByHash.
283 // Returns SDCH_OK if the addition was successfull, and corresponding error
285 SdchProblemCode
AddSdchDictionary(const std::string
& dictionary_text
,
286 const GURL
& dictionary_url
,
287 std::string
* server_hash_p
);
289 // Remove an SDCH dictionary
290 SdchProblemCode
RemoveSdchDictionary(const std::string
& server_hash
);
292 // Registration for events generated by the SDCH subsystem.
293 void AddObserver(SdchObserver
* observer
);
294 void RemoveObserver(SdchObserver
* observer
);
296 static scoped_ptr
<DictionarySet
> CreateEmptyDictionarySetForTesting();
298 // For investigation of http://crbug.com/454198; remove when resolved.
299 base::WeakPtr
<SdchManager
> GetWeakPtr();
302 struct BlacklistInfo
{
303 BlacklistInfo() : count(0), exponential_count(0), reason(SDCH_OK
) {}
305 int count
; // # of times to refuse SDCH advertisement.
306 int exponential_count
; // Current exponential backoff ratchet.
307 SdchProblemCode reason
; // Why domain was blacklisted.
310 typedef std::map
<std::string
, BlacklistInfo
> DomainBlacklistInfo
;
311 typedef std::set
<std::string
> ExperimentSet
;
313 // Determines whether a "Get-Dictionary" header is legal (dictionary
314 // url has appropriate relationship to referrer url) in the SDCH
315 // protocol. Return SDCH_OK if fetch is legal.
316 SdchProblemCode
CanFetchDictionary(const GURL
& referring_url
,
317 const GURL
& dictionary_url
) const;
319 // Support SDCH compression, by advertising in headers.
320 static bool g_sdch_enabled_
;
322 // Support SDCH compression for HTTPS requests and responses. When supported,
323 // HTTPS applicable dictionaries MUST have been acquired securely via HTTPS.
324 static bool g_secure_scheme_supported_
;
326 // A simple implementation of a RFC 3548 "URL safe" base64 encoder.
327 static void UrlSafeBase64Encode(const std::string
& input
,
328 std::string
* output
);
330 DictionaryMap dictionaries_
;
332 // List domains where decode failures have required disabling sdch.
333 DomainBlacklistInfo blacklisted_domains_
;
335 // List of hostnames for which a latency experiment is allowed (because a
336 // round trip test has recently passed).
337 ExperimentSet allow_latency_experiment_
;
339 // Observers that want to be notified of SDCH events.
340 // Assert list is empty on destruction since if there is an observer
341 // that hasn't removed itself from the list, that observer probably
342 // has a reference to the SdchManager.
343 ObserverList
<SdchObserver
, true> observers_
;
345 base::ThreadChecker thread_checker_
;
347 base::WeakPtrFactory
<SdchManager
> factory_
;
349 DISALLOW_COPY_AND_ASSIGN(SdchManager
);
354 #endif // NET_BASE_SDCH_MANAGER_H_